FEBS Letters 586 (2012) 32–35

journal homepage: www.FEBSLetters.org

Accommodating variety in iron-responsive elements: Crystal structure of 1 B IRE bound to iron regulatory protein 1 ⇑ William E. Walden, Anna Selezneva, Karl Volz

Department of Microbiology and Immunology, University of Illinois at Chicago, Chicago, IL 60612, United States article info abstract

Article history: Iron responsive elements (IREs) are short stem-loop structures found in several mRNAs encoding Received 3 October 2011 proteins involved in cellular iron metabolism. Iron regulatory proteins (IRPs) control iron homeo- Revised 14 November 2011 stasis through differential binding to the IREs, accommodating any sequence or structural varia- Accepted 15 November 2011 tions that the IREs may present. Here we report the structure of IRP1 in complex with transferrin Available online 24 November 2011 receptor 1 B (TfR B) IRE, and compare it to the complex with H (Ftn H) IRE. The two IREs Edited by Kaspar Locher are bound to IRP1 through nearly identical protein-RNA contacts, although their stem conforma- tions are significantly different. These results support the view that binding of different IREs with IRP1 depends both on protein and RNA conformational plasticity, adapting to RNA variation while Keywords: Iron metabolism retaining conserved protein-RNA contacts. Iron regulatory protein Ó 2011 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. Iron-responsive element Translation Structure

1. Introduction We have sought to determine the X-ray crystal structures of various IREs bound to IRP1 in an effort to determine how IRE vari- Cellular iron uptake, utilization, and storage are tightly controlled ety is accommodated in this protein:RNA complex. Here we report through the action of iron regulatory proteins 1 and 2 (IRP1 and the crystal structure of IRP1 in complex with IRP2). IRPs bind to iron-responsive elements (IREs) in the non-cod- B (TfR B) IRE, and compare it with the IRP1-bound Ftn H. ing regions of several mRNAs encoding proteins of iron metabolism, regulating message translation or stability in response to cellular 2. Materials and methods iron status [1,2]. To date, nine mRNAs have been confirmed to contain functional IREs [3]. They are all 30 nucleotide stem-loop 2.1. Preparation of protein, RNA, and complex structures with a terminal pseudotriloop (CAGUGX), a five base-pair upper helix, a mid-stem cytosine bulge (C8), and a variable lower The IRP1 protein was of the rabbit (Oryctolagus cuniculus) se- helix. The X-ray structure of ferritin H (Ftn H) IRE in complex with quence, with the double mutation C437S/C503S. These cysteine IRP1 revealed that binding occurs through two distinct protein:IRE substitutions, which do not affect IRE binding, were necessary to regions, primarily involving the conserved C8 and pseudotriloop [4]. suppress protein oxidation and improve homogeneity for single With the exception of the terminal loop and C8, there is little se- crystal growth [9]. The IRE RNA sequence can be considered to also quence conservation among IREs from different mRNAs [5]. Bulged be from rabbit (all known vertebrate TfR B IREs are the same, ex- nucleotides in the stem helices also contribute to IRE variety. These cept for Gallus gallus). A GC base pair was added to the bottom of variable features of IREs dictate that different IRE conformations be the IRE stem for greater stability. The RNA was purchased from accommodated in IRP:IRE complexes, perhaps through unique bind- Dharmacon. Prior to crystallization, the IRP1:TfR B IRE complex ing interactions between IRP and each IRE or the conformational was put in a sample buffer of 20 mM tris, pH 7.5, 5 mM NaCl, plasticity of the IRP. Such differences also are likely to contribute 5 mM DTT, and 0.1 M EDTA, at a concentration of 1.2 mg/ml, based to hierarchical IRP:IRE affinities, providing for differential control on RNA. of the IRE-containing mRNAs by IRPs [6–8]. 2.2. Crystallization, structure determination, and refinement

Abbreviations: IRP, iron regulatory protein; IRE, iron response element; TfR, transferrin receptor 1B; Ftn, ferritin Tetragonal (P41212) crystals of the IRP1:TfR IRE B complex were ⇑ Corresponding author. Fax: +1 312 996 6415. grown in conditions of 0.7 M sodium citrate and 0.1 M HEPES, pH E-mail address: [email protected] (K. Volz).

0014-5793/$36.00 Ó 2011 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.febslet.2011.11.018 W.E. Walden et al. / FEBS Letters 586 (2012) 32–35 33

7.5 at 22 °C. The crystals grew as tetragonal bipyramids. Diffraction described [9]. The internally labelled IRE RNA probes were prepared data were obtained from one crystal at liquid nitrogen temperature with 32P-UTP as done previously [20], and the filter binding assays using X-rays of 1.0000 Å wavelength at the Southeast Regional Col- were performed accordingly. Binding constants were determined laborative Access Team (SER-CAT) 22-ID beamline at the Advanced by RNA saturation under equilibrium conditions. The protein con- Photon Source, Argonne National Laboratory. The data were pro- centrations were held constant at 30 pM, and the RNA concentra- cessed and reduced with the program X-GEN [10]. The structure tions ranged from 2 to 250 pM. The experiments was done in was solved by molecular replacement with the program PHASER triplicate. Kd and Bmax values were calculate from non-linear curve [11] using the published structure of the IRP1:Ftn H IRE complex fits using GraphPad Prizm 4.0b software (GraphPad Software, Inc.).

(PDB 2IPY, [4]), and has been refined to an Rf = 27.5% and Rw = 22.6% at 3.0 Å resolution (Table S1) using Phenix [12] and CNS [13]. The Ramachandran statistics (RAMPAGE, [14]) show 3. Results and discussion 90.7% of backbone angles in the favored region, 8.0% in the allowed, and 1.3% in the outlier regions (Fig. S1). The final structure contains Fig. 1A shows the predicted secondary structure of the TfR B IRE 95% of all atoms in the complex plus 146 solvent molecules. All in comparison with that of Ftn H. The two IREs share the canonical nucleotides of the IRE are present, as well as all residues of the IRE stem-loop structure but differ significantly in sequence and IRP1 molecule except for the N-terminal His-tag, and 40 residues composition at the inter-helical hinge. Crystals of the IRP1:TfR B in the three unresolved loops spanning residues 126–146, 500– IRE complex were obtained and the structure was solved by molec- 511, and 623–629. ular replacement as described in Materials and Methods. The IRP1:Ftn H IRE complex (old PDB ID 2IPY) was re-refined The IRP1 protein adopts the bilobal, L-shape conformation, with during this project after it was discovered that the two protein mol- domain positions and local loop conformations in the IRP1:TfR B ecules in the asymmetric unit (previously refined independently) IRE complex the same as seen earlier with the Ftn H IRE (Figs. S2 had very high non-crystallographic symmetry (NCS). Refinement and S3). The most striking differences between the two structures with tight NCS restraints on the protein gave a 3% reduction in Rf (fi- are in the stems of the bound IREs (Fig. 1B). The helices are hinged nal: Rf = 21.8% and Rw = 19.8%) and improved bond geometry for as rigid bodies, connected through a bend near base pair 7–25. The both protein and RNA (new PDB ID 3SNP). The two RNA molecules bend angle of the bound TfR B IRE is only 8°, while the IRP1-bound in the asymmetric unit did not obey the NCS. Ftn H IRE is kinked with a bend of 20° [4]1 plus an extra twist of 9°. The different bends in the bound IREs lead to changes of up to 2.3. Structural interpretation 8 Å in the approach of the IRE lower helix to domain 4 of IRP1 (Fig. 2). This slightly alters the set of protein:RNA contacts (Table Least squares superpositions of protein molecules were done 1). For example, Arg 688, which makes three H-bonds with the with LSQKAB of CCP4 [15] and COOT [16] using all available Ca Ftn H IRE lower helix is too far away from the lower helix of TfR atoms. Relative domain positions were analyzed with the program B IRE to make the same contacts. In fact, the side chain of Arg DynDom [17]. The overall conformation of the IRP1 protein in the 688 is poorly resolved in the IRP1:TfR B IRE complex. On the other two complexes is the same (Figs. S2 and S3). There are slight shifts hand, a potential for interaction of the lower helix with Arg 704 in IRP1 domains 3 and 4, but the differences are not measurably was gained in the complex. Thus, the overall number of potential significant, and the directions of displacement correlate with inter- bonds observed in the two complexes is essentially equivalent, molecular contacts, suggesting minimal packing effects. Unex- consistent with the relative binding affinities that we observed plainable localized differences in the protein occur in one area for these IREs with IRP1 (KD of 34 ± 16 pM for TfR B and involving residues 174–176, 205–207, and 539–542. This area is 45 ± 19 pM for Ftn H IREs; Fig. S6; see also [6,7,21]). near the 430 and 530 loops that are important for IRP1 conforma- The difference in stem-loop bend angles for the IRP1-bound TfR tional switching and ligand binding. B and Ftn H IREs can be attributed to the composition of the interh- Superpositions of RNA molecules were done with LSQKAB of elical hinges. The TfR B IRE has a more shallow bend angle because CCP4 [15] and COOT [16] using identical nucleotides (e.g., C8, it (like most non-ferritin IREs) is missing the unpaired U6 of the A12, C14–G18, and U21 for upper helix and loop) plus equivalent evolutionarily ancestral ferritin form [5]. Absence of the U6 bulge ribose-phosphate atoms when appropriate (e.g., remaining back- permits the axes of the upper and lower helices to be more coax- bone of upper helix). Measurements of the inter-helical bend an- ially in-line, forming a relatively uninterrupted A-form helix for gles were calculated with the program 3DNA [18] from the dot the entire stem. This demonstrates that the single bulged cytosine products of the global linear helical axes of the lower and upper has just a minor effect on the helical course of the full-length stem helices, each defined by their five base pairs. The two crystallo- (see also [22]). The more shallow bend angle of the TfR B IRE actu- 1 graphically independent Ftn H RNA molecules had bend angles ally closes up the major groove, and rotates the lower helix away of 23.1° and 17.7°. All angles have estimated uncertainties of ±3°. from the stem-binding domain (domain 4) of the IRP1 molecule Solvent accessibility and buried surface areas were calculated (Fig. 1B), as discussed above. with the program PISA [19]. The areas of the buried protein:RNA The upper helix of the stem holds the C8 bulge and terminal interfaces for the two crystallographically independent IRP1:Ftn loop at the appropriate spacing and orientation to allow insertion 2 H IRE complexes were 1434 and 1413 Å , and that for the IRP1:TfR in their respective binding pockets. There is no sequence conserva- 2 2 B IRE complex was 1442 Å . tion between the TfR B and Ftn H IRE upper helices, only structural conservation of the five base pair A-form helix geometry. It was re- 2.4. Binding assays cently reported [8] that sequence variation in the upper helix, particularly at the closing base pairs, could affect affinity of an Protein-RNA affinities were measured by nitrocellulose filter- IRE for IRP1. Given the similarities in the structures of the IRP1- binding assays. Protein was expressed and purified as previously bound TfR B and Ftn H IREs in this region, and the lack of se- quence-specific contacts between IRP1 and the IRE upper helix, 1 Some recalculated bend angles do not agree with the original reports because of the impact of sequence differences in the upper helix on protein the different methods or programs used. For consistency, all angles here are binding may relate to effects on helical twist and pitch, and/or to calculated the same way. 2 Coordinates and diffraction data have been deposited in the Protein Data Bank as effects on helix stability. entries 3SN2 for the IRP1:TfR B IRE and 3SNP for the IRP1:Ftn H IRE. 34 W.E. Walden et al. / FEBS Letters 586 (2012) 32–35

Fig. 1. IREs compared in this study. (A) Secondary structures of transferrin receptor 1 B and ferritin H IREs. The GC base pair at the bottom of the TfR B IRE (grey box) was introduced for stability. The outlined regions have the same three-dimensional structures. (B) Superimposed IRP-bound TfR B (dark) and Ftn H (light) IREs. Superposition was based on equivalent atoms in the outlined regions in A. See the Supporting Information for details.

Fig. 2. Differences in positions of phosphorus atoms of IRP1-complexed TfR B IRE and the two IRP1-complexed Ftn H IREs after superposition of upper helices, C8 bulges, and loops.

IRP1 recognizes the common structural features of the TfR B and IRP1 in the crystal structure [4]. This suggests that IRP1:IRE Ftn H IREs in binding: C8 and the terminal loop. The protein:RNA interactions promoted by the presence of an unpaired residue at bonding patterns involving those groups are essentially the same position 6, such as those with the lower helix, contribute to overall (Table 1). Deviations were seen in the vicinity of the loop, most binding affinity of ferritin IREs. notably at variable nucleotide 19. Bulged C19 of the TfR B IRE The sequence and structural variation seen among IREs raises adopted a well-ordered conformation in the complex, whereas the question of whether other functions, such as binding with the U19 base of the Ftn H IRE had no electron density. The course other proteins, targeting the mRNA for degradation, or differential of the RNA backbone in this region (residue 19–22) also differs interaction with IRP1 and IRP2 [7] might drive the character of (Fig. 2). Remarkably, only minor bonding differences were seen these regulatory RNA elements. In this regard, recent findings sug- in this region for the two complexes, although five potential con- gest that ferritin IREs themselves specifically bind Fe2+, and that tacts to the RNA backbone occur here (Table 1). It seems likely that this destabilizes the IRP1:IRE complex [23]. Interestingly, deletion this reflects the induced fit and conformational plasticity of the IR- of U6 largely abrogated this effect of Fe2+ on IRP1:Ftn H IRE inter- P1:IRE interaction. action. Mitochondrial IRE, which lacks an unpaired U6, Given that the main points of contact between IREs and IRP1 are also was more refractory to the effect of Fe2+. Since there is high se- the terminal loop and bulged C8, the significance of variation in quence conservation among IREs of an mRNA family (e.g., ferritin elements below C8—such as the presence or absence of an mRNAs), it is intriguing to consider that IREs and their interaction unpaired U6 and the orientation of the lower helix relative to the patterns with IRPs have each evolved to uniquely respond to regu- upper helix of the IRE—is presently unknown. Deletion of U6 in fer- latory signals for control of iron metabolism. ritin IREs lowers the binding affinity for IRPs approximately 3-fold There are as yet no general rules regarding RNA stem-loop [7,8]. U6 of the Ftn H IRE itself has no apparent interaction with bending in protein:RNA recognition: some bulged RNA stem-loops W.E. Walden et al. / FEBS Letters 586 (2012) 32–35 35

Table 1 References Comparison of protein:RNA H-Bond distances.

IRP1 IRE dist. (Å) Ftn Ha dist. (Å) TfR B [1] Wallander, M., Leibold, E.A. and Eisenstein, R.S. (2006) Molecular control of vertebrate iron homeostasis by iron regulatory proteins. Biochim. Biophys. Interactions near terminal loop Acta 1763, 668–689. His207 NE2 U19/C19 O2P 3.6 3.2 [2] Volz, K. (2008) The functional duality of iron regulatory protein 1. Curr. Opin. Arg269 NH1/2 U17 O4 3.4 2.7 Struct. Biol. 18, 106–111. Asn298 ND2 G18 O1P 3.5 4.7 [3] Wang, J. and Pantopoulos, K. (2011) Regulation of cellular iron metabolism. Asn298 ND2 G18 O2P 3.5 4.0 Biochem. J. 434, 365–381. Glu302 OE1 U17 O20 2.7 2.5 [4] Walden, W.E., Selezneva, A.I., Dupuy, J., Volbeda, A., Fontecilla-Camps, J.C., Ser371 OG A15 N6 3.5 3.4 Theil, E.C. and Volz, K. (2006) Structure of dual function iron regulatory protein 1 complexed with ferritin IRE-RNA. Science 314, 1903–1909. Ser371 OG A15 N7 2.6 2.9 [5] Piccinelli, P. and Samuelsson, T. (2007) Evolution of the iron-responsive Lys379 NZ G16 O6 3.3 3.1 element. RNA 13, 952–966. Thr438 OG1 U17 O1P 2.4 3.0 [6] Guo, B., Yu, Y. and Leibold, E.A. (1994) Iron regulates cytoplasmic levels of a Asn439 ND2 U17 O2P 3.0 3.1 novel iron-responsive element-binding protein without aconitase activity. J. 0 Asn535 O C15 O2 3.0 2.8 Biol. Chem. 269, 24252–24260. Asn535 ND2 G16 O2P 2.6 3.0 [7] Ke, Y., Wu, J., Leibold, E.A., Walden, W.E. and Theil, E.C. (1998) Loops and Arg536 NH2 U20/C20 O20 3.1 2.8 bulge/loops in iron-responsive element isoforms influence iron regulatory Arg536 NH2 U20/C20 O30 3.4 3.3 protein binding. J. Biol. Chem. 273, 23637–23640. [8] Goforth, J.B., Anderson, S., Nizzi, C.P. and Eisenstein, R.S. (2010) Multiple Interactions near C8 and hinge determinants within iron-responsive elements dictate iron regulatory protein Ser681 OG C8 N4 3.2 3.1 binding and regulatory hierarchy. RNA 16, 154–169. Pro682 O C8 N4 2.7 2.8 [9] Selezneva, A., Cavigiolio, G., Theil, E.T., Walden, W. and Volz, K. (2006) Gly684N C8 O1P 2.9 3.3 Crystallization and preliminary X-ray diffraction analysis of iron regulatory Gly710N C8 O1P 2.9 2.9 protein 1 in complex with ferritin IRE RNA. Acta Cryst. F62, 249–252. Ser708 OG C8 O2P 2.7 2.6 [10] Howard, A.J. (2000) Data processing in macromolecular crystallography. Ser778 OG U10/G10 O1P 3.5 3.2 Chapter in: Crystallographic Computing 7: Proceedings from the Asp781N C8 O2 3.2 2.9 Macromolecular Crystallographic Computing School, 1996, Oxford Arg780 NH1 U9/G9 O1P 3.4 3.2 University Press, P. E. Bourne and K. D. Watenpaugh, eds. Oxford. [11] McCoy, A.J. et al. (2007) Phaser crystallographic software. J. Appl. Cryst. 40, Interactions with lower helix 658–674. b Asn685 OD1 G26/U26 N2 2.7 – [12] Adams, P.D. et al. (2010) PHENIX: a comprehensive Python-based system for c Arg688 NE A28 O1P 2.9 – macromolecular structure solution. Acta Cryst. D66, 213–221. Arg688 NH2 A28 O1P 3.2 –c [13] Brunger, A.T. et al. (1998) Crystallography and NMR System: A new software Arg688 NH2 A29/U29 O2P 2.8 –c suite for macromolecular structure determination. Acta Cryst. D54, 905–921. Arg704 NE A29/U29 O1P 5.1 3.6 [14] Lovell, S.C. et al. (2002) Structure validation by Ca geometry: u/w and Cb Arg704 NH2 C30 O1P 6.3 3.6 deviation. Proteins, Structure, Function and Genetics 50, 437–450. Arg728 NH2 C25/A25 O20 3.4 2.8 [15] CCP4: COLLABORATIVE COMPUTATIONAL PROJECT, NUMBER 4 (1994) The CCP4 Suite: Programs for Protein Crystallography. Acta Cryst. D 50, 760–763. a Averages from the two complexes in the asymmetric unit. [16] Emsley, P. and Cowtan, K. (2004) Coot: model-building tools for molecular b No N2 because of U at position 26. graphics. Acta Cryst. D60, 2126–2132. c No Arg688 side chain density in IRP1:TfR B IRE complex. [17] Hayward, S. and Berendsen, H.J.C. (1998) Systematic analysis of domain motion in proteins from conformational change: New results on citrate synthase and T4 lysozyme. Proteins, Structure, Function and Genetics 30, 144– exhibit dynamic stem flexion in solution, while others are rigid 154. [24–26]; reviewed in [25]. The solution structure of an early IRE [18] Lu, X.-J. and Olson, W.K. (2008) 3DNA: a versatile, integrated software system model was a semi-rigid, two-helix hairpin with no unpaired U6 for the analysis, rebuilding and visualization of three-dimensional nucleic- that had an interhelical bend of 9° [27].1 That model is most sim- acid structures. Nat. Protoc. 3, 1213–1227. [19] Krissinel, E. and Henrick, K. (2007) Inference of macromolecular assemblies ilar to the TfR B IRE in this study, with its bend angle of 8°. The re- from crystalline state. J. Mol. Biol. 372, 774–797. sults here show that interhelical bending by itself may not be a [20] Swenson, G.R., Patino, M.M., Beck, M.M., Gaffield, L. and Walden, W.E. (1991) significant determination of the strength of IRP1:IRE binding. How- Characteristics of the interaction of the ferritin repressor protein with the iron responsive element. Biol. Met. 4, 48–55. ever, IRE stem-loop flexion in solution could still be a selection pro- [21] Leibold, E.A., Laudano, A. and Yu, Y. (1990) Structural requirements of iron- cess that precedes complex formation. This is consistent with the responsive elements for binding of the protein involved in both transferrin recent proposal that hierarchical IRP1:IRE binding could be accom- receptor and ferritin mRNA post-transcriptional regulation. Nucleic Acids Res. 18, 1819–1824. plished through IRE differences that affect the initial mechanism of [22] Xiong, Y. and Sundarlingam, M. (2000) Two crystal forms of helix II of Xenopus IRE recognition [8]. laevis 5S rRNA with a cytosine bulge. RNA 6, 1316–1324. [23] Kahn, M.A., Walden, W.E., Goss, D.J. and Theil, E.C. (2009) Direct Fe+2 sensing by iron-responsive messenger RNA-repressor complexes weakens binding. J. Acknowledgements Biol. Chem. 284, 30122–30128. [24] Finger, L.D., Trantirek, L., Johansson, C. and Feigon, J. (2003) Solution structures We are grateful to the staff of Southeast Regional Collaborative of stem-loop that bind to the two N-terminal RNA-binding domains of nucleolin. Nucleic Acids Res. 31, 6461–6472. Access Team (SER-CAT) at the Advanced Photon Source, Argonne [25] Getz, M., Sun, X., Casiano-Negroni, A., Zhang, Q. and Al-Hashimi, H.M. (2007) National Laboratory for their assistance and support. Use of the Ad- NMR studies of RNA dynamics and structural plasticity using NMR residual vanced Photon Source was supported by the Department of Energy dipolar couplings. Biopolymers 86, 384–402. under Contract No. W-31-109-Eng-38. This work was funded by [26] Dethoff, E.A., Hansen, A.L., Musselman, C., Watt, E.D., Andricioaei, I. and Al- Hashimi, H.M. (2008) Characterizing complex dynamics in the transactivation the National Institutes of Health (GM-71504 to K.V.). response element apical loop and motional correlations with the bulge by NMR, molecular dynamics, and mutagenesis. Biophys. J. 95, 3906–3915. [27] McCallum, S.A. and Pardi, A. (2003) Refined solution structure of the iron- Appendix A. Supplementary data responsive element RNA using residual dipolar couplings. J. Mol. Biol. 326, 1037–1050. Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.febslet.2011.11.018.